Developing Understanding of Robot Minds by Kimberly A. Brink A
Total Page：16
File Type：pdf, Size：1020Kb
A Theory of (the Technological) Mind: Developing Understanding of Robot Minds by Kimberly A. Brink A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy (Psychology) in The University of Michigan 2018 Doctoral committee: Professor Henry M. Wellman, Chair Professor Margaret Evans Professor Susan Gelman, Professor Benjamin Kuipers Kimberly A. Brink [email protected] ORCID iD: 0000-0001-5947-2350 Kimberly A. Brink 2018 DEDICATION This dissertation is dedicated to the strong and wonderful women in my life who were there when I needed them. Thank you, Tyler, Emma, Margaret, Jenn, Shee Shee, Lisa, Steph, Katie, and April. Thank you, Ann and Heather. And thank you to my mom. ii ACKNOWLEDGMENTS This dissertation would not exist if not for Henry M. Wellman, my primary advisor, who entertained and nurtured my many interests. I will forever be indebted to him. I thank my committee members Margaret Evans, Susan Gelman, and Benjamin Kuipers for their invaluable feedback and investment in both me and my work. I also thank my collaborator, Kurt Gray, for his contributions to this work. This research would not have been possible without the funding and training provided by the National Institute of Child Health and Human Development, the Rackham Graduate School, and the Center for Human Growth and Development. Though this is my dissertation, the research reported here reflects the contributions of many research assistants with whom I have had the great pleasure to work. I thank those research assistants who dedicated their time and efforts to assist with these projects as well as related projects not reported on here. Countless thanks go to: Clare Anderson, Kaela Brown, Annie Dreisbach, Lorene Dreisbach, Makenzie Flynn, Kristen Goncalves, Sara Guttentag, Taryn Hayes, Alyssa Krentzel, Emily Kunster, Alysse Quinn, Caitlin Schneider, Rebecca Seguin-Skrabucha, Kayla Winer, Alexis Yezbick, and Lauren Yousif. A special thanks also goes to the laboratory coordinators who made this work possible: Cassie Neely, Hope Peskin-Shepherd, and Talie Toomajian. In addition to the many hours invested by my research assistants, the research reported on here would not have been possible without the support of the University of Michigan Living Lab iii Program. Thank you to Craig E. Smith for bringing the model to Ann Arbor. Thank you, also, to the University of Michigan Museum of Natural History (Kira Berman, Brittany Burgess) for generously supporting my work. Outside the lab, I received excellent teaching mentorship from Drs. Ioulia Kovelman, Shelly Schreier, Rich Gonzalez, Daniel Weissman, and Matthew Diemer. From them, I learned not only how to teach, but how to mentor students and research assistants in ways that would promote their learning and nurture their curiosities. Thank you for helping me be a better teacher and researcher. Throughout my time at Michigan, I also benefited from many great friendships. I thank my cohort for their helpful feedback and constant encouragement: Sammy Ahmed, Nkemka Anyiwo, Emma Beyers-Carlson, Fernanda Cross, Margaret Echelbarger, Arianna Gard, Amira Halawah, Tyler Hein, Jasmine Manalel, Lauren Tighe, and Neelima Wagley. I also thank my lab mates, past and present: Maria Arredondo, Lindsay Bowman, Amanda Brandone, Jasmine DeJesus, Rachel Fine, Selin Gulgoz, Danielle Labotka, Jonathan D. Lane, Kristan Marchak, Rebecca Marks, Ariana Orvell, Rachna Reddy and Steven Roberts. Beyond my cohort and the lab, I thank Paige Safyer, Jennifer Dibbern, Shen Huang, Shee Shee Jin, Lisa Johnson, Stephanie Oh, Jyotsna Venkataramanan, Katie Asher Halquist, Maggie Halpern, Nicholas Alderink, and April Rogers for their friendship. To even get to this point, I am heavily indebted to several people who led me to this path. First, I thank Rebecca Saxe for igniting my passion for psychological research. Second, I thank Laura Schulz for introducing me to the incredible world of infant research. Third, I thank Luca Bonatti for taking a chance on a research assistant whom he’d never met to cross an ocean and work with babies. Fourth, I thank Heather Ann Thompson for inspiring and supporting my iv pursuit of graduate school. Fifth, I thank Dillon Thompson Erb for guiding and supporting me toward this path. And sixth, I thank Henry M. Wellman, once more, for allowing me to pursue a novel and odd research topic with endless enthusiasm and steadfast support. Lastly, I thank those closest to me who were there when times were light and dark. Thank you to Tyler, Jenn, and Katie, the incredible women who were on the other side of a phone call, cup of coffee, or glass of wine whenever I needed them. And thank you to my parents, Marty and Lisa Brink, and my big brothers, Matt and Nathan Brink, who all nurtured and encouraged my eccentricities. I could not have asked for a more supportive, loving, and wonderful family. v TABLE OF CONTENTS DEDICATION ii ACKNOWLEDGMENTS iii LIST OF TABLES viii LIST OF FIGURES ix LIST OF APPENDICES x ABSTRACT xi CHAPTER I. Introduction 1 Children’s Developing Understanding of Robots 4 Research Questions 7 Outline of Research 8 II. Creepiness Creeps In: Uncanny Valley Feelings Are Acquired in Childhood 9 Abstract 9 Method 15 Results 21 Discussion vi III. Robot Teachers for Children? Young Children Trust Robots Depending on their Perceived Accuracy and Agency 36 Abstract 36 Study 1 39 Study 2 48 General Discussion 53 IV. A Theory of the Robot Mind: Developing Beliefs about the Minds of Social Robots 58 Abstract 58 Study 1: Adults 64 Study 2: Children 78 General Discussion 89 V. General Discussion 96 Related Research on Robot Instructors 98 Future Directions 100 REFERENCES 105 vii LIST OF TABLES TABLE II.1 Exploratory Factor Analysis of Interview Items 20 II.2 Regression Analyses Predicting Uncanniness Difference Scores 22 III.1 Social Robot Responses for Each Trial 42 III.2 Inanimate Machine Responses for Each Trial 51 IV.1 Proportion of Responses in Each Type of Mental Attribution 73 viii LIST OF FIGURES FIGURE I.1 An Array of Robots Designed Specifically to Interact with Children 4 II.1 A Schematic Depiction of the Theoretical Uncanny Valley 11 II.2 Still Frames from the Videos of Each Robot 13 II.3 Images to Assist Children in Answering Likert-Scale Type Questions 17 II.4 The Interaction Between Robot Type and Age 23 II.5 Measure of Uncanniness 25 II.6 Creepy-Weird Stimuli Used in Follow-Up Interview 27 III.1 Still Frame from Study 1 41 III.2 Familiar and Novel Objects Shown in Both Study 1 and Study 2 43 III.3 Proportion of Correct Responses to Accuracy, Ask, and Endorse Questions 46 III.4 Still Frame from Study 2 50 IV.1 Still Images of 10 Robots Presented in Study 1 62 IV.2 A Diagram of the CFA Model for Studies 1 and 2 69 IV.3 A Visualization of K-Means Clustering for Adult Responses 72 IV.4 The Additional Robot ASIMO, Included in Study 2 79 IV.5 A Visualization of K-Means Clustering for Child Responses 84 ix LIST OF APPENDICES APPENDIX II.A Robot Beliefs Interview 33 III.A Reduced Robot Beliefs Interview 56 IV.A Robot Beliefs Interview – Adults 93 IV.B Robot Beliefs Interview – Children 94 x ABSTRACT The purpose of this dissertation is to explore how children attribute minds to social robots and the impacts that these attributions have on children’s interactions with robots, specifically their feelings toward and willingness to trust them. These are important areas of study as robots become increasingly present in children’s lives. The research was designed to address a variety of questions regarding children’s willingness to attribute mental abilities to robots: (1) To what extent do children perceive that social robots share similarities with people and to what extent do they believe they have human- like minds? (2) Do attributions of human-like qualities to robots affect children’s ability to understand and interact with them? (3) Does this understanding influence children’s willingness to accept information from robots? And, of crucial importance, (4) how do answers to these questions vary with age? Across a series of five studies, I investigated children’s beliefs about the minds of robots, and for comparison adults’ beliefs, using survey methods and video stimuli. Children watched videos of real-life robots and in response to targeted questions reported on their beliefs about the minds of those robots, their feelings about those robots, and their willingness to trust information received from those robots. Using a variety of statistical methods (e.g., factor analysis, regression modeling, clustering methods, and linear mixed-effects modeling), I uncovered how attributions of a human-like mind impact feelings toward robots, and trust in information received from xi robots. Furthermore, I explored how the design of the robot and features of the child relate to attributions of mind to robots. First and foremost, I found that children are willing to attribute human-like mental abilities to robots, but these attributions decline with age. Moreover, attributions of mind are linked to feelings toward robots: Young children prefer robots that appear to have human-like minds, but this reverses with age because older children and adults do not (Chapter II). Young children are also willing to trust a previously accurate robot informant and mistrust a previously inaccurate one, much like they would with accurate and inaccurate human informants, when they believe that the robot has mental abilities related to psychological agency (Chapter III). Finally, while qualities of the robot, like behavior and appearance, are linked to attributions of mind to the robot, individual differences across children and adults are likely the primary mechanisms that explain how and when children and adults attribute mental abilities to robots (Chapter IV).